Location: Remote (US/EU time-zone overlap required)
Type: Full-time
Reports to: Board of Directors
The role:
UEX.US is a cryptocurrency exchange shipping Trading Pro, a savings/rewards product, and a payments stack. We move fast, and we ship. What we don't yet have is one person who owns the answer to: Is this product actually good, is it actually competitive, and does it actually do what we told users it does?
That's this job. You are the user's advocate inside the building and the quality gate for the devs.
This is an AI-first role: We are not hiring a product manager who occasionally uses ChatGPT. We are hiring someone who has rebuilt their entire working method around AI and produces the output of a four-person product team. If that isn't already how you work, this role will be uncomfortable, and you should not apply.
What you'll own:
- User feedback as a system. Discord, support tickets, app-store reviews, churn conversations — pulled in continuously, tagged, quantified by AI pipelines you build and maintain, turned into a ranked list nobody has to argue about. Anecdote becomes evidence, weekly, without a research team.
- Competitive intelligence, continuously. Binance, Kraken, Bybit, Coinbase, MEXC, Gate. Fees, spreads, listings, UX flows, onboarding friction, what they ship weekly. Automated monitoring and AI-generated teardowns — you should know what our competitors changed this week before they announce it.
- Test everything we already have. Every existing feature, hands-on, repeatedly — deposits, withdrawals, order types, KYC, rewards accrual, mobile. AI-assisted test scripts and regression checks where they help; your own hands where they don't. You find the breakage before users do.
- Promise integrity. Every claim on the site, in the app, in the ToS and in support macros must match what the product actually does. Where it doesn't, you flag it and drive it to zero. This is a regulated money services business — a UI label that overstates what a feature does is not a copy problem; it's an exposure.
- Priority discipline. You set the roadmap with the board, then you make it real: specs the devs can build from, acceptance criteria, sprint follow-through, and honest weekly status. Shipped-and-verified, not shipped.
- Metrics. Define them, instrument them, report them. Activation, retention, volume per active trader, support-ticket rate per feature.
## AI requirements — the hard bar
This section carries the most weight in our evaluation. You must be able to demonstrate, live, in the interview:
- Daily production use of frontier models (Claude, GPT, Gemini) as your primary working environment — not as a novelty, as the place your work happens.
- Prompt engineering at a professional level: structured prompts, XML/JSON output contracts, few-shot examples, chain-of-thought, evaluating and iterating on prompt quality rather than accepting first output.
- Building repeatable AI workflows, not one-off chats. Projects, custom instructions, saved prompt libraries, reusable analysis templates that a teammate could run.
- AI over large unstructured datasets: feeding thousands of support tickets, Discord messages and reviews into a pipeline that returns a defensible, quantified priority list — and knowing how to check that it didn't hallucinate the numbers.
- Agentic and tool-using AI: Claude Code, Cowork, MCP connectors, browser agents, or equivalent. Comfortable letting an agent run a multi-step task and reviewing its work.
- API-level literacy. You can call a model API directly, batch a job, chain calls, and estimate what a workflow costs before you run it. Light scripting (Python or JS) to glue things together.
- AI-assisted data work: natural-language querying over product data, generating SQL and validating it, building dashboards without waiting on an engineer.
- AI for spec and QA production: drafting PRDs, acceptance criteria, edge-case enumeration, test matrices, and user-f
